Cannabis-derived products (flowers, infused butter, edibles, creams, ointments, water-based products, glycerine-based products, essential oil concentrates (waxes, shatters, budders, etc.) are all typically "heterogeneous" (not completely evenly mixed). For any heterogeneous sample (i.e., almost all cannabis-derived samples other than ethanol tinctures at lower concentrations), pulling different samples from the same batch can yield slightly different results. The key to reproducible results that are representative of the batch is in "sampling technique" and the measurement of multiple samples from the batch. (See our FAQ on "sampling" for details specific to different types of product.)
When we make a measurement on a submitted sample, the lab techs also perform sampling - i.e., we take several small pieces/parts from different regions of the sample submitted for analysis. For flowers, for example, if the sample is several buds, we will take pieces from many buds, some from bud tops, some from bud bottom, some from interior, some from exterior. We try to take a "representative sample," which means the totality of what we test should have about the same overall composition of the entire sample submitted.
We also take care not to change the material in any way from what we received - with one exception: samples that are to be shaken before use will be shaken before sampling. Otherwise, we do not heat, combine multiple components, or otherwise alter the submitted samples. Thus, it is incumbent upon the client to make sure their own sampling of the batch is also representative of the entire batch. Sample heterogeneity and inadequate client sampling is the biggest source of variation in single lab testing reproducibility. Two different buds from the same plant can vary by a factor of 2 in concentration of cannabinoids and terpenes, for example, especially for indoor grows where lighting of plant tops is significantly greater than lighting of plant bottoms, but significant variations are even observed for plants grown outdoors. Thus, if we get one bud from Client X's Durban Poison on Monday and a second single bud from the same batch on Friday, the chances that those two buds will have the same exact content is nearly zero. Often, clients attribute those differences to the lab, but there's no way around biodiversity short of good sampling and multiple measurements.